Performance Expectations of Branded Autonomous …...Performance Expectations of Branded Autonomous Vehicles: Measuring Brand Trust Using Pathfinder Associative Networks by Natalie
Post on 04-Aug-2020
3 Views
Preview:
Transcript
Performance Expectations of Branded Autonomous Vehicles:
Measuring Brand Trust Using Pathfinder Associative Networks
by
Natalie Celmer
A Thesis Presented in Partial Fulfillment of the Requirements for the Degree
Master of Science
Approved November 2018 by the Graduate Supervisory Committee:
Russell Branaghan, Co-Chair
Erin Chiou, Co-Chair Nancy Cooke
ARIZONA STATE UNIVERSITY
December 2018
i
ABSTRACT
Future autonomous vehicle systems will be diverse in design and functionality
since they will be produced by different brands. In the automotive industry,
trustworthiness of a vehicle is closely tied to its perceived safety. Trust involves
dependence on another agent in an uncertain situation. Perceptions of system safety,
trustworthiness, and performance are important because they guide people’s behavior
towards automation. Specifically, these perceptions impact how reliant people believe
they can be on the system to do a certain task. Over or under reliance can be a concern for
safety because they involve the person allocating tasks between themselves and the
system in inappropriate ways. If a person trusts a brand they may also believe the brand’s
technology will keep them safe. The present study measured brand trust associations and
performance expectations for safety between twelve different automobile brands using an
online survey.
The literature and results of the present study suggest perceived trustworthiness
for safety of the automation and the brand of the automation, could together impact trust.
Results revelated that brands closely related to the trust-based attributes, Confidence,
Secure, Integrity, and Trustworthiness were expected to produce autonomous vehicle
technology that performs in a safer way. While, brands more related to the trust-based
attributes Harmful, Deceptive, Underhanded, Suspicious, Beware, and Familiar were
expected to produce autonomous vehicle technology that performs in a less safe way.
These findings contribute to both the fields of Human Automation Interaction and
Consumer Psychology. Typically, brands and automation are discussed separately
however, this work suggests an important relationship may exist. A deeper understanding
ii
of brand trust as it relates to autonomous vehicles can help producers understand
potential for over or under reliance and create safer systems that help users calibrate trust
appropriately. Considering the impact on safety, more research should be conducted to
explore brand trust and expectations for performance between various brands.
iii
TABLE OF CONTENTS
Page
LIST OF TABLES .......................................................................................................... v
LIST OF FIGURES ....................................................................................................... vi
CHAPTER
1 INTRODUCTION ......................................................................................... 1
2 LITERATURE REVIEW ............................................................................... 4
Trust in Automation ........................................................................... 5
Over and Under Reliance .................................................................... 7
Brands ................................................................................................ 8
Brand Trust and Automation ..............................................................11
Measuring Trust in Automation ..........................................................14
Pathfinder ..........................................................................................15
The Present Study ..............................................................................17
3 METHOD ..................................................................................................... 19
Participants ........................................................................................19
Study Design and Materials ................................................................20
Procedure ..........................................................................................21
4 RESULTS .................................................................................................... 25
Expected Performance Measures ........................................................25
Brand Trust Association Networs .......................................................29
Correlation.........................................................................................32
5 DISCUSSION ............................................................................................... 35
iv
CHAPTER Page
Overview ...........................................................................................35
Limitations and Future Work ..............................................................35
Conclusions .......................................................................................37
REFERENCES .............................................................................................................40
APPENDIX
A AMAZON MECHANICAL TURK SURVEY LINK ....................................... 43
B INFORMED CONSENT ................................................................................. 45
C RANKING ACTIVITY .................................................................................. 48
D RATING ACTIVITY ...................................................................................... 51
E DEMOGRAPHICS QUESTIONAAIRE AND END OF SURVEY .................. 53
F FRIEDMAN’S TEST (WITH WILCOXON POST TESTS) ............................ 56
G RANK DESCRIPTIVE STATISTICS .............................................................. 60
H RANK 1 DESCRIPTIONS: SUMMARY TABLE ........................................... 67
I NUMBER OF NODES BETWEEN EACH BRAND & ATTRIBUTES.............69
J FULL CORRELATION MATRIX .................................................................. 71
v
LIST OF TABLES
Table Page
1. Median (IQR) and Mean (SD) Ranked Position................................................25
vi
LIST OF FIGURES
Figure Page
1. Brand Personality Framword ............................................................................ 9
2. Pathfinder Network Example.......................................................................... 16
3. Histograms - Frequency of Ranked Position ................................................... 26
4. Significant Differences in Ranked Position ..................................................... 27
5. Pathfinder Network ....................................................................................... 31
6. Nearest Neighbor Network ............................................................................ 32
1
CHAPTER 1
INTRODUCTION
Rapid technological innovation introduces great uncertainty in how people should
interact with technology (Van Geenhuizen & Nijkamp, 2003). It is often difficult to
predict how new systems will perform, therefore, people often do not interact
appropriately with technology and this can impact safety Trust in automation is important
because it influences how much people accept and rely on the automation, it influences
people’s behavior toward a system (Lee & Moray 1992). However, with autonomous
vehicles, the automation is also tied to a brand name. Autonomous vehicle technology is
becoming increasingly complex and difficult to understand and although there is a lot left
to learn, we still see these systems begin to populate the media and our roadways.
Advanced autonomous vehicle technology is becoming more variable than existing
automobile technology, such as, automatic gear shifting and traditional cruise control
settings. These older features may also be presented with different interfaces across
brands, however, they function similarly. Especially, in the developmental stages of more
advanced features, we begin to see different interfaces and different functionality.
Consider a pedestrian waiting to cross a street; will a fully autonomous vehicle
stop for them? How will the vehicle communicate to the pedestrian that they may cross?
What will the pedestrian expect the car to do?
The answers may depend on the brand of the vehicle, and a person’s perceptions
of the branded vehicle’s capabilities. One brand of vehicle might always stop, whereas
another might only stop if the pedestrian is a specific distance from the curb. Many
technology corporations and automobile manufacturers are developing autonomous
2
vehicles. The sheer variety of companies with different technologies and design
approaches is likely to yield great diversity. Separate corporations have a common goal to
design, produce, and sell vehicles with autonomous technology. However, to differentiate
vehicles in such a competitive market, systems consist of various features and
programming that represent the different brands they are associated with, including, the
brand’s personality, identity, target consumer groups, and other products and services
they have on the market. The most salient difference is often between interfaces, what the
system looks and feels like. However, sometimes the underlying functionality of similar
feature may actually differ between brands of autonomous vehicles as well. In short,
different brands produce different experiences. It is possible these brand differences yield
different levels of trust in the automation, therefore different expectations for vehicle
performance.
In fact, this situation exists already. Park-Assist (BMW) and Autopark (Tesla) are
both autonomous parking features, however, their Human-Machine Interfaces (HMI) are
different and require different inputs from the person using the feature. For instance,
Tesla employs a streamlined process; about three actions are required to parallel park the
vehicle using Autopark. The experience with BMW, however, is more cognitively
involved. The driver must initiate the Park-Assist feature, press the brake, turn on their
blinker, read a pop-up message stating they understand that they are liable for the
vehicle’s ultimate performance, confirm this by pressing OK, then they must press and
hold the Park Assist button for the duration of the entire parking process, and release the
brake when the process is complete.
3
Safety guidelines and standard requirements do not remedy these inconsistencies.
The U.S. Department of Transportation and the National Highway Traffic Safety
Administration (NHTSA) provide guidelines in Automated Driving Systems 2.0: A Vision
for Safety (September 2017). However, these standards still allow producers great
freedom in implementation. For example, one guideline states: “HMI design should also
consider the need to communicate information regarding the Automated Driving
System’s state of operation relevant to the various interactions it may encounter and how
this information should be communicated” (pg. 10). Just as a rubric for an academic
assignment does not lead students to submit identical projects, the NHTSA guidelines
address broad safety concerns and leave room for variety in system designs and
configurations.
When considering autonomous vehicles, the technical capabilities of the
automation and brand associations may both contribute to expectations and trust for the
system. The follow review of the literature considers the potential influence of branding
on user perceptions of – and expectations for – the safety of autonomous vehicles. relates
principles of Cognitive and Social Psychology, Marketing, Consumer Psychology, and
Human-Automation Interaction, identifies the gaps within the research, and seeks to
explore relationships between brand trust and trust in automation.
4
CHAPTER 2
LITERATURE REVIEW
A literature search was conducted using various databases including Google
Scholar, PsycINFO and Academic Search Premiere via the ASU Library Catalog.
Searches included varying combinations of key words including, “trust”, “brand trust”,
“brand personality”, “branding”, “associations”, “automation”, “automation bias”,
“autonomy”, “autonomous vehicles”, “HAI”, “trustworthiness” and “safety”.
Articles for this review were selected from various journals in the areas of
Marketing Research, Social Psychology, Consumer Psychology, Human Factors, and
Human-Automation Interaction. Federal and public sources provided by NHTSA were
also referenced. Many brand-related articles report findings of brand trust, brand affect
and brand loyalty related to predicting purchasing behavior. However, studies and articles
that primarily focused on branding and price or purchasing decisions we excluded from
the review. Additionally, many of the articles related to trust in automation focused on
measuring trust in the moment or after interacting with a system. In the literature, less
focus tended to be placed on prospective trust, or trust expectations, prior to observing or
experiencing the system’s performance. Though, there are many factors that are reported
to influence trust. This review focuses primarily on automation and autonomous vehicles.
Therefore, articles related to trust, expectations and safety judgements of automation as it
may relate to an associated brand are found in this review. Since this is a broad topic, a
large number of articles were scanned, and 38 sources are referred to in this review.
Broad ideas include cognitive biases, brands, trust in automation, automation bias, brand
trust, self-congruity and risk judgements.
5
Trust in Automation
Trust in automation is the belief that another agent will help in uncertain, or
vulnerable, situations (Lee & See, 2004). Trust is often goal oriented. For autonomous
vehicle systems, a primary goal is to travel from point A to point B in a safe, efficient,
and pleasant way. Trust is based on the expectation that when given control, the system is
capable of performing, and even improving, the driving task while most importantly,
keeping the passengers safe. Trust depends on how successful the person expects the
automation to be (Lee & Moray 1992; Sheridan 1992; Lee & See 2004). This expectation
guides a person’s behavior with a system (Mosier, Skitka, & Heers, 1998; Lee & See,
2004).
The amount of trust a person has in a system should reflect the system’s
capabilities, especially when monitoring and occasional intervention are required. For
example, when a driver must switch between an Autopilot feature and manual control.
Otherwise, when trust is not appropriately calibrated, human-automation systems often
break down (Lee & See, 2004). When a system breaks down it is not working as
intended, this typically suggests poor performance can potentially cause harm.
Various design characteristics affect expectations and trust in autonomous
systems (Lee & See, 2004). Choi and Ji (2015) explored factors that influence trust in
autonomous vehicles specifically. The goal of this study was to explore factors that
influence trust in automation, and how a person’s level of trust in the system can predict
the likelihood they will adopt and accept autonomous vehicles. They surveyed 552
drivers and discovered three constructs to positively impact trust. These factors included,
system transparency, technical competence, and situation management. They found trust
6
perceptions tend to increase for highly transparent systems, more technically competent
systems, and systems with acceptable situation management (Choi & Ji, 2015). They also
found increased trust in the system led to decreased perceived risk. Increased trust was
also shown to increase the likelihood that the person would adopt and accept autonomous
vehicles.
These factors are closely tied to system reliability, which is the degree to which
the automation does what it is intended to do. Previously, most work involving human-
automation interaction did not focus on dynamic, real-world environments. However,
Desai et al. (2012) conducted a study to mimic real-world, unstructured situations in
which autonomy reliability is not as stable. This study explored the effects of fluctuating
reliability on trust in automation and use of its capabilities. Decreases in system
reliability were shown to decrease trust (Desai et al., 2012). Results also indicated that
increased trust in automation was linked to low perceived risk and low cognitive load
(Desai et al., 2012). This supports the idea that system designs should address situational
characteristics such as, perceived risks, workload, and task difficulty. These factors and
trust seem to mutually reinforce one another. However, expectations for system
functionality, true system capability, and the person’s role are often mismatched.
An example of this is Automation Bias; when a person favors the use of
automation over their own input. This often results from an over trusting attitude and
leads to an inappropriate level of reliance on the system (Mosier, Skitka, & Heers, 1998).
Other instances resulting from inappropriate calibrations of trust include Misuse, Disuse
and Abuse (Parasuraman & Riley, 1997). In 1997, Parasuraman and Riley synthesized
theoretical, empirical, and analytical work regarding human use, misuse, disuse, and
7
abuse of automation technology. They defined each of these instances and claimed that a
deeper understanding of these use cases will inspire improvements in system design,
training, and policies regarding the use of automation.
Over and Under Reliance
When a person thinks the system is capable of things that it actually is not capable
of, they tend to over-trust the system. It often results in misuse of, or over reliance on, the
system (Parasuraman & Riley, 1997). This trust-based behavior can compromise safety.
For example, consider Tesla’s Autopilot feature. People who think they can be less
involved in the driving task are misusing the system. Based on the vehicle’s true
capabilities, drivers are still required to be vigilant, to supervise the vehicle in the driving
task. People who have the impression that they can watch videos or send text messages
instead of monitoring the vehicle, are over-relying on the system. These individuals are
under the impression that the system is capable of things that it actually is not capable of.
Therefore, they may over-trust the system, and allocate tasks to the system that are more
complex than it is able to handle. In our Autopilot example, the person may give up all
control and depend entirely on the automation to drive down a road while they text on
their phone. In the event that the vehicle comes across an obstacle or situation it doesn’t
recognize. This could lead to poor performance ranging from a near crash to even a fatal
accident.
Contrastingly, low levels of trust can promote disuse, which is under-reliance on
the system (Parasuraman & Riley, 1997). A driver who refuses to use any autonomous
features in a vehicle is under-relying on the system. This may occur because the person
doubts the vehicle’s technological capabilities, or they believe they are better. This is can
8
also be a safety concern because human drivers do not always perform well. There may
be times where the automation is safer and performs better than a human could. For
example, an automated braking feature may detect an obstacle in a blind spot and stop the
vehicle just before impact. Autonomous features in vehicles are intended to improve
driver and vehicle performance and safety. They may not be perfect, but they are
intended to promote safer and more efficient driving than a human driver alone (Beiker,
2012).
Traditionally, in the literature, it is often the case that autonomous vehicles and
various types of advanced technology are discussed in isolation – without regard to
environmental factors. But ultimately, people don’t actually experience or interact with
technology in that way. Instead, we have a ton of real-world information, like a brand for
example, that may affect our perceptions.
Brands
A brand is a name, term, or symbol that distinguishes a seller’s product or service
from others (Bennett, 1995). An important part of branding entails the accumulation of
associations and perceptions, in memory linked to a brand (Aaker, 1991). In essence, it
includes what consumers know (or believe) about products. Branding and associations
affect people’s behaviors and interactions with products and services (Rossiter & Percy,
1991).
According to Deighton (1992), brands “promise a future performance”. They set
expectations for the quality of their product (Keller, 1993). Trust in the brand is
established through fulfillment of these expectations over time (Delgado-Ballester, 2003).
From the moment the brand is born, it is associated with specific values, limitations, and
9
target consumer groups (Kotler & Andreasen, 1991). Brands are also closely linked to the
performance and quality of their products and services (Keller, 1993; Zeithaml 1988),
specifically, how reliable and successful the product is at fulfilling its intended purpose.
In relation to a brand, reliability and trustworthiness perceptions are an individual’s belief
a specific brand will perform in a particular, or positive, way.
Additionally, people tend to spontaneously ascribe human personality
characteristics to brands, creating a brand personality that summarizes brand associations.
To categorize these personalities, Jennifer Aaker (1997) developed an empirically
derived framework (figure 1) of five dimensions of brand personality, and 15 associated
facets, shown below. These dimensions, Sincerity, Excitement, Competence,
Sophistication, and Ruggedness are useful for describing and summarizing brand
associations.
This brand personality framework was based on Malhotra’s (1981) work with
construct scales. Originally, Aaker started with a list of 309 traits. She reduced this by
more than half and then told participants to rate 37 brands on these traits. After numerous
trials, and a factor analysis – five dimensions of brand personality emerged; including,
Figure 1: Brand Personality Framework (Aaker, 1997)
10
sincerity, excitement, competence, sophistication, and ruggedness. From these
dimensions, she identified and defined 15 associated facets. Each facet adds detail to one
of the five dimensions and describes it by providing context. For example, the
competence dimension is supported by three facets (reliable, intelligent, successful).
These brand personality dimensions are useful for describing and summarizing
brand associations. For example, it might be appropriate for a good brand of autonomous
vehicle to be closely associated with a competent brand personality. Since competence
and reliability are closely tied to trustworthiness (Wojiciszke & Abele, 2008; McCroskey
& Teven, 1999; Fiske, Cuddy & Glick, 2007); which translates to safety and the ability to
perform in the automotive industry.
Additionally, the dimensions and facets are also useful for differentiating brands
from one another (Freeling & Forbs, 2005). Brand exposure often evokes strong,
automatic, and subconscious inclinations and feelings about a product (Thomson et al.,
2005). Considering the pedestrian example, one brand of autonomous vehicle may be
designed to stop for the pedestrian to cross in a way that ensures the pedestrian that it is
okay to do so, while another may not. The brand personalities of each may help the
pedestrian decide to walk or not. In any interaction with autonomous vehicles, a person
may simply base their trust in the system on associations. They may take specific actions
surrounding an autonomous vehicle based on its brand personality.
Brand Trust and Automation
Trust is a fundamental component of good relationships; it evolves based on past
experience (Rempel et al., 1985; Rotter, 1980). Interpersonal trust commonly discussed
in the literature as how much one is willing to accept vulnerability, or risk. It is
11
determined by a perception of the world and the likelihood that others, or the
environment, would harm the self (Rotter, 1980, Robinson, Shaver, & Wrightsman,
1991). Regardless of individual differences in interpersonal trust, or willingness to trust
others, people identify patterns in intentions, behaviors, motivations, and qualities linked
to a positive outcome (Rotter, 1980; Rempel et al., 1985). Trust is not dichotomous. It is
not simply a matter of trust or distrust, instead, levels of trust fall along a continuum.
Brand trust is the level of security associated with a brand. It is based on the
perceived reliability of the brand, and how responsible it is for the welfare of the
consumer (Delgado-Ballester, 2003). Brand trust is also context dependent. It is specific
to the nature of the situation and the other agents involved (Mayer et al., 1995; Schaefer
et al., 2016). Trust-based relationships between consumers and brands, resemble that of
humans and automation. Similarly, human-automation trust is based on expectations of
system capabilities.
History-based trust focuses on past performance (Merritt & Ilgen, 2008), and how
it relates to future interactions. Brands form relationships with consumers by meeting, or
exceeding, their expectations. In this way, brands build trust by being predictable (Mayer
et al., 1995), providing good experiences time after time.
Automation is designed to build trust in the same way. A trustworthy system is
simple and understandable. It acts in the operator’s best interest, is designed to induce
proper trust calibration, shows performance history and meets the operator’s performance
expectations (Lee & See, 2004). Autonomous vehicle systems produced by different
brands will vary in these characteristics.
12
For instance, Carlson et al. (2013) demonstrated that trust in a vehicle’s capabilities
was higher for autonomous vehicles created by a well-known brand than for an unknown
brand. This work identified factors that influence trust in branded autonomy, such as,
statistics of past performance, extent of research on the car’s reliability, predictability,
credibility of the engineers, technical capabilities, and possibility for system failure.
Carlson et al. (2013) examined factors that influence trust in two domains;
autonomous vehicles and medical diagnosis systems, and within two dimensions; safety
criticality and brand recognizably. In this study, participants ranked 29 factors based on
their influence on trust. In the autonomous vehicle domain, statistics of the car’s past
performance ranked the highest. The extent of research on the car’s reliability and
credibility of the engineers who designed the car were also ranked within the top six
factors.
Therefore, it is not surprising that there was a significant difference between trust in
systems produced by a well-known company and systems produced by an unknown
brand, or small start-up company (Carlson et al., 2013). Results indicated that participants
trusted the vehicles capabilities more when it was created by a well-known brand,
Google. Participants rated the statements: “I trust the machines’ capabilities because it
was created by Google”, and, “My trust in a fully-autonomous system similar to this
machine would decrease if it was created by a lesser- known company.” These findings
were reinforced between different groups that were asked the questions in the opposite
direction. Higher trust in the capabilities of technology created by well-known companies
than a lesser known company was shown in both domains (autonomous vehicles and
13
medical diagnosis systems), in both safety-critical and non-safety-critical situations
Carlson et al. (2013).
Results demonstrated that brand associations influence trust in autonomous
technology. In this study, past performance, reliability, predictability, technical
capabilities, and credibility of the engineers emerged as top influential factors for trust
(Carlson et al., 2013). These are similar to the factors proposed by Lee and See (2004)
that include, performance, process, and purpose. Where performance is how well the
automation completes the task, process is a person’s experience with the system and their
opinion for how it works, and purpose is the system’s intention. A deeper understanding
of the similarities and differences of these two models of trust, in automation and in
brands, can how they influence trust in human-autonomous vehicle can help produce
safer and more desirable systems. These insights can inform autonomous vehicle
producers design decisions. For instance, if it is known that people tend to think a certain
brand of vehicle will be more safe, or more capable of controlling a vehicle on the road,
then designers and marketers can present the automation in a way that promotes
appropriate trust calibration, and prevents over or under reliant behavior.
Brand is an element of autonomous systems that is often excluded in the
exploration of trust in the realm of human-automation interaction. As mentioned above,
trust in autonomous vehicles was shown to be higher for well-known brands than lesser
known brands (Carlson et al, 2013), however it would be valuable to know if, and how,
trust in automation varies between various well-known brands.
14
Measuring Trust in Automation
Generally, trust in automation is largely dependent on performance expectations;
how successful the person expects the automation to be (Lee & Moray 1992; Sheridan,
1992; Lee & See 2004). Trust in automation has been characterized and measured across
numerous domains. For example, Singh, Molloy, and Parasuraman (1993) used factor
analysis to develop a Complacency-Potential Rating scale that measures automation
induced complacent behaviors. Much like Automation Bias, discussed above,
complacency is when a person is overconfident in the system, it tends to be associated
with over reliant behaviors. Complacency-potential was shown to vary depending on a
person’s trust and confidence in the automation, in addition to their ultimate reliance on
automation. Singh, Molloy, and Parasuraman (1993) demonstrated that complacency-
potential can be measured using ratings that capture general attitude towards items related
to everyday automation technology.
In proposing a quantitative model of trust, Sheridan (1988) also suggested seven
attributes of trust for which systems vary. These included, reliability, robustness,
familiarity, understandability, explication of intention, usefulness and dependence.
Additionally, Jian, Bisantz, and Drury (2000) developed an empirically derived
scale to measure trust in automation. Using a series of experimental phases, they
compared words relating to trust across three types of trust-based relationships; trust
between people, trust between people and automation, and trust in general. Experimental
phases included a word elicitation phase, a questionnaire phase, and a paired comparison
phase. A cluster analysis was used to identify twelve factors, or attributes, related to trust
between people and automation. Their results also suggested that trust and distrust are
15
opposites, not separate concepts. Therefore, these attributes include words related to both
trust and distrust. The twelve attributes were Deceptive, Underhanded, Suspicious,
Beware, Harmful, Confidence, Security, Integrity, Dependable, Reliable, Trustworthy,
and Familiarity. Further, Jian, Bisantz, and Drury (2000) used these twelve factors to
develop a scale to measure a participant’s trust in an autonomous system. The scale
consists of twelve statements, such as, “The system is deceptive”, and participants are
instructed to rate each statement using a 7-point scale with 1 representing, “Not at All”
and 7, “Extremely”. Essentially, the scale is a series of Likert-style questions that
measure how similar the participants impression of the system is to the statement.
Pathfinder
A Pathfinder algorithm (Schvaneveldt et al., 1989; Schvaneveldt, 1990) is a
quantitative tool that can be used with pairwise relatedness data to create network models
that illustrate associations, or similarities (Branaghan & Hildebrand, 2011). Pairwise
relatedness data provides insight for how each of the comparison items are interrelated. It
can be collected with multiple pairwise comparison tasks. For instance, to compare a list
of many items, one could ask participants to consider two items at a time and rate their
relatedness, or similarity, on a scale. Participants would do this until each item on the list
has been compared against all others, so that every possible paired combination is rated.
Given this type of data a pathfinder algorithm is used to link more related items visually
using a node. The resulting network structure is a visual depiction of the perceived
relatedness between the comparison items. The organization and structure of the network
is outlined by nodes that link each of the comparison items together. The distance, or
number of nodes, between comparison items in the network represents their relatedness.
16
In the network structure, highly related items are directly connected by a single node. For
less related items, however, one or more links may be present between the comparison
items. For example, if concept A and concept C are very related then a single node would
connect them. Similarly, if concept C and concept B are very related then a single node
would also connect them to one another. If concept A and concept B are less related, they
would only be connected through their relationship to concept C, so two nodes would
stand between them (see figure 2 below).
Figure 2: Pathfinder Network Example
Nodes within a pathfinder networks can also be quantified with weighted values.
Higher weight values indicate that items are more related, lower weight values indicate
items are less related.
Branaghan and Hildebrand (2011) used a Pathfinder algorithm to create visual
networks for the relationship between a participant’s self-image and certain brands. In
their study, a pairwise comparison task was used to collect relatedness data between
brand personality and automobiles. Participants also compared their self-image with the
15 dimensions and facets of Aaker’s Brand Personality Framework (1997). They also
compared 12 automobiles in this way. The pathfinder algorithm measured the match
between each item compared to the others. The distance between the items represented
their relatedness, these values were used to create association networks. These
relationships were represented using associative networks because: “brand personality
17
and self-image are knowledge structures, and facets of brand personality also describe the
self” (Branaghan & Hildebrand, 2011, p. 304). Not only did the resulting networks
illustrate congruity between the self and brand personalities, but also illustrated
associations between the other brands included in the study. Branaghan and Hildebrand
(2011) also demonstrated that brand personality associations are related to brand
preferences.
The Present Study
The aim of the present study was to investigate the relationship between brand
trust associations and performance expectations for safety of autonomous vehicles
produced by various brands.
Previous work has shown trust in autonomous vehicles produced by well-known
brands was higher than trust in autonomous vehicles produced by a lesser-known brand
(Carlson et al., 2013). The present study however, explores trust across various well-
known brands. Because increased trust in automation is linked to low perceived risk
(Desai et al., 2012), it was hypothesized that brands more closely related to trustworthy
attributes will be ranked as more safe. Therefore, brand trust associations would be
positively correlated with performance expectations for safety. Such that, the closer a
brand is associated with trust (opposed to distrust) the better a person would expect an
autonomous vehicle produced by that brand to perform. Additionally, the further a brand
is associated with distrust (opposed to trust) the worse a person would expect an
autonomous vehicle produced by that brand to perform. Meaning, on average participants
would predict better performance for safety for the brands more associated with
18
trustworthiness, and worse performance for safety for the brands less associated with
trustworthiness.
19
CHAPTER 3
METHOD
Participants
Participants were recruited online using the Amazon Mechanical Turk
(www.mturk.com) crowdsourcing tool for survey data collection. All participants were
compensated $1.00US for their participation, upon full completion of the survey. This
amount was used on Amazon Mechanical Turk because it encouraged participation in the
online survey without impacting the participants financial situation. One hundred and
three participants were recruited for this study. This sample size was based on the sample
size used in previous studies (Jian, Bisantz, & Drury, 2000; Carlson et al., 2013). Of the
one hundred and three participants, the data from four participants was eliminated
because it was deemed inaccurate. Inaccurate data was defined as a survey completed in
3 minutes or less, as this would be too fast to accurately read through each survey
question. Additionally, inaccurate data also included surveys containing contradicting or
incomplete responses.
This study included Mechanical Turk users with worker accounts who had
participated in at least previous 50 tasks and maintained a 95% or above Human
Intelligence Task (HIT) approval rating; this was done to encourage reliable data
collection (Paolacci & Chandler, 2014). Since materials were presented in the English
language, participants also had Amazon Turk worker accounts with a registered location
within the United States.
At the end of the survey, participants completed a demographics questionnaire
(Appendix E). Of the 99 participants, 61 were male, 36 were female, 1 identified as other,
20
and 1 chose not to provide this information. All participants in this study were ages 18
and older. 50 of the 99 participants were between the ages of 18 and 34, 43 were between
the ages of 35 and 64, 5 were 65 years old or older and 1 chose not to provide this
information. Educational backgrounds varied, but all participants had at least a High
School diploma and the majority of participants (60 of 99) completed an Associates,
Bachelor’s, or Master’s degree.
Additionally, 91 were licensed drivers and 85 owned their own vehicle. This
sample of 85 participants consisted of 13 Toyota owners, 12 Honda owners, 10 Ford
owners, 7 Nissan owners, 5 BMW owners, 5 Chevrolet owners, 5 Mazda owners, 2
Volvo owners, and 26 other owners (brands included, Hyundai, KIA, Saturn, Mitsubishi,
Dodge, Saturn, Jeep, Infinity, Audi, Range Rover, Mercury, Lexus, Cadillac, GMC,
Subaru, and Suzuki).
Study Design and Materials
The present exploratory research study intended to investigate the relationship
between two variables, brand-trust associations and performance expectations for
branded autonomous vehicles. Brand-trust associations are characteristics or attributes
related to trustworthiness, linked to a specific brand. Performance expectations are the
predicted outcomes of branded autonomous vehicles when executing an action.
In order to measure these two variables, a two-part survey was used. All
participants responded to the online survey using Qualtrics, which they accessed through
their Amazon Worker account. One part of the survey included a ranking activity, the
other included Likert-style rating questions.
21
The twelve automobile brands selected for this survey included, Volvo, Mercedes,
Volkswagen, BMW, Toyota, Honda, Nissan, Mazda, Ford, Chevy, Chrysler, and Tesla.
The twelve brands vary in their target consumer groups and average vehicle price.
Generally, European, American, and Japanese brands were equally represented.
Additionally, brands owned by the same corporation were not included (e.g., since
Chrysler was included, Jeep was not).
The twelve trust-based attributes included in this study were, Deceptive,
Underhanded, Suspicious, Beware, Harmful, Confidence, Feeling secure, Integrity,
Dependable, Reliable, Trustworthy, and Familiarity. These attributes were selected and
adapted from the empirically developed and validated scale for trust in autonomous
systems (Jian, Bisantz, & Drury, 2000).
Procedure
Participants volunteered to participate in the study online by selecting the survey
from their list of open surveys on their Amazon Mechanical Turk Worker’s Account. On
average it took participants 23 minutes to complete the survey.
First, participants read a brief description of this survey, including time
commitment and monetary compensation amount, and clicked a link to participate
(Appendix A). This link opened a new browser winder with the Qualtrics Survey.
Participants received a brief introduction encouraging them to take their time and then
they were provided an informed consent form (Appendix B).
The Ranking Activity: Expected Performance for Safety. After informed consent
was obtained, a ranking activity, was used to collect expected performance data for each
of the twelve brands (Appendix C). Participants were provided the following scenario:
22
“Imagine you are a passenger in a driverless car. It is fully autonomous, meaning there is
no need for a human driver. This car was designed to drive itself on the road and operate
in the same environments and conditions that a person could. You are traveling down the
road and you see a pedestrian crossing the street in front of you. You believe the vehicle
needs to stop for them.”
Then participants were asked to evaluate and rank the twelve brands based on
their expectation for how an autonomous vehicle produced by each brand would perform
in the above scenario. Participants ranked the automobile brands from 1 to 12, where 1
represented the vehicle that would be the most safe for themselves, the pedestrian, and all
others on the road and 12 represented the least safe for themselves, the pedestrian, and all
others on the road. These rankings were based on their current knowledge and
expectations. The listed of automobile brands was presented in a randomized order for
each participant.
The ranking measure was used because we were interested in how the provided
list of brands compare to one another, on average. Though scoring each brand
individually on a scale would have provided similar information, ranking encourages the
incorporation of underlying associations and latent perceptions of the twelve brands in
this particular evaluation.
As a supplement, participants were asked to provide a couple words of their own
to describe the brand they ranked as most safe (1) in the ranking activity. This served a
dual purpose because the free response format helped to identify participants who were
providing inaccurate data and the qualitative responses provided insight for how people
determined their responses. For example, one person who selected BMW as most safe
23
(rank 1) stated: “It's a brand that I haven't heard of any problems with in a very long time.
I would trust this brand the most to be able to have safe and functioning products.”.
The Rating Activity: Brand-Trust Associations. Then, participants completed
Likert-style rating questions to measure relatedness between each automobile brand and
each trust-based attribute, which as previously stated, we defined as brand-trust
associations.
For each of the twelve brands, in randomized order, participants responded to the
following question, “How related is <BRAND NAME> to each of the following?”.
Twelve trust-based attributes were listed beneath this, and participants provided a
relatedness rating for each attribute using a 7-point scale, where a score of 1 indicated
Not Related At All and a score of 7 indicated Extremely Related (Appendix D). The list of
trust-based attributes was presented in a randomized order for each participant.
In total, participants completed 144 pairwise comparison ratings which informed
the relatedness data later used to quantify associations between each automotive brand
and each trust-based attribute. The structure of this survey was adapted from the pairwise
comparison task used by Branaghan and Hildebrand (2011). However, instead of
participants comparing their self-image and 12 automobiles with the 15 dimensions and
facets of Aaker’s Brand Personality Framework (1997), participants in the present study
compared 12 automobile brands and the 12 attributes of Jian, Bisantz, and Drury’s scale
to measure trust in automation (2000).
Finally, participations completed a demographics questionnaire to collect
information regarding their gender, age, educational background, and automobile
24
ownership. Three questions to measure components of interpersonal trust, brand
preference behavior, and brand trust were also included (Appendix E).
Upon completion, participants were thanked for sharing their opinions, notified
their participation was concluded, and Amazon Mechanical Turk would facilitate the
compensation process by transfer their earnings to their account. This message included a
unique survey code. Participants returned to the Amazon Mechanical Turk portal to enter
their unique survey code into the space provided (Appendix A). Participants received
their compensation within one to three days after submitting their unique survey code.
No identifying information was associated or linked to any individual responses.
The Qualtrics survey was set up to ensure that IP Addresses were not recorded. The
unique survey codes were the only link to a worker’s ID, however, it was only used to
approve survey completion and distribute compensation, it was deleted immediately after
compensation was distributed.
25
CHAPTER 4
RESULTS
Expected Performance Measures
The ranking activity required all participants to put all twelve brands in order
from most safe to least safe (1 to 12 respectively) based on their expectation for how an
autonomous vehicle produced by each brand would perform. Since ranked position is an
ordinal variable, the Friedman’s test was conducted (Appendix F) as a non-parametric
alternative to a repeated measures one-way ANOVA. The Friedman’s test does not
assume normally distributed samples. Therefore, it was used to determine whether there
was a significant difference in ranked position between brands.
The Friedman’s test revealed that there was a significant difference in ranked
position between brands, χ2(11) = 246.3, p < 0.00. Median (IQR) and Mean (SD) ranked
positions of each brand are listed in the table below.
Automobile Brand Median Mean (SD)
BMW 3 4.07 (2.90)
Mercedes 3 4.23 (2.90)
Tesla 3 4.64 (4.00)
Volvo 4 5.24 (3.26)
Toyota 6 5.92 (2.95)
Honda 6 6.14 (3.29)
Volkswagen 6 6.72 (3.12)
Ford 9 7.67 (3.30)
Nissan 8 7.75 (2.64)
Chevrolet 9 8.14 (2.85)
Mazda 9 8.57 (2.56)
Chrysler 9 8.92 (2.57)
Table 1. Median (IQR) and Mean (SD) ranked position
26
The test statistic of the Friedman’s test is called the Friedman’s Q and is notated
with Chi-square. This test statistic represents and summarizes how far the average ranks
are from one another and to what degree does the average explains the variance. Similar
to the idea of variance, Friedman’s Q would be zero if the mean ranks were equal to one
another and would increase as the mean ranks become further apart.
The Friedman’s Test is an omnibus test that revealed a significant difference
between typical ranked position, however, it cannot identify which brands differed from
one another. However, the frequency distributions for ranked position are shown for each
brand in figure 3. The histograms are listed in order by mean ranked position. Therefore,
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
Rank
Freq
uenc
y
Tesla H
onda Volvo
Mercedes
Volkswagen
BMW
ToyotaN
issanM
azdaFord
Chevrolet
Chrysler
1 2 3 4 5 6 7 8 9 10 11 12
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
Rank
Freq
uenc
y
Tesla H
onda Volvo
Mercedes
Volkswagen
BMW
ToyotaN
issanM
azdaFord
Chevrolet
Chrysler
1 2 3 4 5 6 7 8 9 10 11 12
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
Rank
Freq
uenc
y
Tesla H
onda Volvo
Mercedes
Volkswagen
BMW
ToyotaN
issanM
azdaFord
Chevrolet
Chrysler
1 2 3 4 5 6 7 8 9 10 11 12
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
Rank
Freq
uenc
y
Tesla H
onda Volvo
Mercedes
Volkswagen
BMW
ToyotaN
issanM
azdaFord
Chevrolet
Chrysler
1 2 3 4 5 6 7 8 9 10 11 12
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
Rank
Freq
uenc
y
Tesla H
onda Volvo
Mercedes
Volkswagen
BMW
ToyotaN
issanM
azdaFord
Chevrolet
Chrysler
1 2 3 4 5 6 7 8 9 10 11 12
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
Rank
Freq
uenc
y
Tesla H
onda Volvo
Mercedes
Volkswagen
BMW
ToyotaN
issanM
azdaFord
Chevrolet
Chrysler
1 2 3 4 5 6 7 8 9 10 11 12
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
Rank
Freq
uenc
y
Tesla H
onda Volvo
Mercedes
Volkswagen
BMW
ToyotaN
issanM
azdaFord
Chevrolet
Chrysler
1 2 3 4 5 6 7 8 9 10 11 12
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
Rank
Freq
uenc
y
Tesla H
onda Volvo
Mercedes
Volkswagen
BMW
ToyotaN
issanM
azdaFord
Chevrolet
Chrysler
1 2 3 4 5 6 7 8 9 10 11 12
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
Rank
Freq
uenc
y
Tesla H
onda Volvo
Mercedes
Volkswagen
BMW
ToyotaN
issanM
azdaFord
Chevrolet
Chrysler
1 2 3 4 5 6 7 8 9 10 11 12
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
Rank
Freq
uenc
y
Tesla H
onda Volvo
Mercedes
Volkswagen
BMW
ToyotaN
issanM
azdaFord
Chevrolet
Chrysler
1 2 3 4 5 6 7 8 9 10 11 12
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
Rank
Freq
uenc
y
Tesla H
onda Volvo
Mercedes
Volkswagen
BMW
ToyotaN
issanM
azdaFord
Chevrolet
Chrysler
1 2 3 4 5 6 7 8 9 10 11 12
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
TeslaHonda
VolvoMercedes
VolkswagenBM
WToyota
Nissan
Mazda
FordChevrolet
Chrysler
0.0 2.5 5.0 7.5 10.0 12.5
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
010203040
Rank
Frequency
Rank
Freq
uenc
y
Tesla H
onda Volvo
Mercedes
Volkswagen
BMW
ToyotaN
issanM
azdaFord
Chevrolet
Chrysler
1 2 3 4 5 6 7 8 9 10 11 12
Freq
uenc
y
Figure 3: Histograms - Ranked Distributions
27
brands listed at the top of the chart tended to be ranked as more safe, and brands listed
towards the bottom of the chart tended to be ranked as least safe (1 to 12 respectively).
Though the frequency distributions in figure 3 demonstrate a general trend, in
order to explicitly determine which brands significantly differed from one another, the
Wilcoxon Signed-Ranked test is needed for all 66 paired combinations. This is a non-
parametric alternative to a paired-samples t-test.
A post hoc analysis conducted with Wilcoxon Signed-Ranked test and Bonferroni
correction, was conducted to determine the effect of brand name on expected
performance for safety (appendix F). The test revealed 36 significant differences in
average ranked position between brand pairs. The remaining 30 pairs did not significantly
differ from one another.
Figure 4 summarizes these findings. Figure 4 shows the groups brands that did not
significantly differ from one another and divides brands that did significantly differ. For
example, BMW (in red) is significantly different from all 8 brands outside of the red
BMW
Volkswagen
Nissan
Chrysler
Chevy
Ford
Honda
Toyota
Volvo
Tesla
Mercedes
Mazda
Figure 4: Significant differences in Ranked Position between brands
28
bracket and not significantly different from the 3 brands inside the red bracket (Mercedes,
Tesla, Volvo). Furthermore, Volkswagen (in purple) is significantly different from all 5
brands outside of the red bracket and not significantly different from the 6 brands inside
the red bracket (Volvo, Toyota, Honda, Ford, Nissan, Chevy).
Another component of the expected performance measure (ranking activity), was
a supplementary free response question that asked participants to provide a couple of
words to describe the brand they ranked as most safe (1 of 12). Therefore, counts were
gathered for each brand based on the number of participants who ranked the brand as
most safe (1 of 12). Of the 99 participants, 37 ranked Tesla, 15 ranked Volvo, 14 ranked
BMW, 13 ranked Mercedes, 8 ranked Honda, 5 ranked Toyota, 4 ranked Ford, 1 ranked
Nissan, 1 ranked Chevrolet, 1 ranked Chrysler, while no participants ranked Mazda nor
Volkswagen as most safe (1 of 12).
A qualitative data analysis was conducted to identify themes in the brand
descriptions. Within the responses, six major themes emerged. Brands that participants
expected to perform most safe in the pedestrian example were noted to, be the first brand
to be successful in this space of automobile technology, be advanced and innovative
brands in the automotive industry, in general, be high quality, or luxury automobile
brands, have a good reputation for safety, be consistently reliable, and functional, in
general.
Of the 37 participants who expected Tesla to be the most safe, 12 attributed this to
their belief that Tesla was the first brand to be successful in this space of automobile
technology and 15 attributed this to their belief that Tesla is advanced and innovative
brands in the automotive industry, in general. Of the 15 participants who expected Volvo
29
to be the most safe, 14 attributed this to their belief that Volvo has a good reputation for
safety. Of the 14 participants who expected BMW to be the most safe, participants were
split. These participants thought BMW to be advanced and innovative brand in the
automotive industry, a high quality, or luxury automobile brand, and consistently reliable.
Of the 13 participants who expected Mercedes to be the most safe, 10 attributed this to
their belief that Mercedes is a high quality, luxury brand. A summary table containing all
counts for each theme facetted by brand can be found in Appendix H.
Brand-Trust Association Networks
The relatedness data from the brand trust rating activity in the survey was used to
construct a Pathfinder network. Since, the survey collected relatedness ratings for all 12
brands and all 12 attributes, for a total of 144 attribute-brand ratings, the pairwise
comparison ratings were first translated in Mat Lab using a scaling method to account for
all 276 possible combination pairs consisting of attribute-brand, attribute-attribute, and
brand-brand paired combinations.
This scaling method was used to derive relatedness measures for each brand-
brand pair by using the Pearson Product Moment to correlate attribute ratings for each
brand with each other brand. Similarly, trait-trait relatedness measures were calculated by
correlating the automobile ratings for each attribute with every other trait. All relatedness
measures were scaled by subtracting the minimum score on each scale from each
individual score and then dividing that score by the maximum score on the scale. This
translated the data to a normalized scale ranging from 0 to 1. The resulting data was
combined into a n x n relatedness matrix and used as proximity data for the brand trust
30
association network. This proximity data was used to calculate the distance between all
twelve brands and all twelve trust-based attributes.
The Pathfinder Algorithm tool (downloaded at http://interlinkinc.net/index.html)
was used to average all ratings, create a visual network of brand-trust associations, and
explore patterns of brand-trust associations. The resulting network illustrates the
underlying relationship patterns based on perceived relatedness of the twelve trust-based
attributes and twelve brand names (see figure 4 below).
The network in figure 4 illustrates that brand trust associations and relative
perceived trustworthiness tended to vary between automobile brands. For instance, BMW
and Mercedes are directly associated to Confidence, one node connects them. Confidence
is directly linked to Trustworthiness. BMW and Mercedes are also most disassociated
with the attributes representing distrust.
Conversely, the network shows that Volkswagen is most closely related to
distrusting attributes, such as, Deceptive, Underhanded, Harmful, Suspicious, and
Beware, and most disassociated with Confidence and Trustworthiness. A summary table
listing the number nodes between each item in the network can be found in Appendix I.
Similar to a benchmarking study where automobile manufacturers are interested
in identifying how their brand compares to their competitors, this analysis provides
insight for relative trust of individual brands.
31
To further investigate this relationship, the brand-trust relatedness data was
transformed to examine the direct relatedness between brands, for example, “How related
is <Tesla> to <Chrysler>”. This was done because the twelve trust-based attributes
initially developed by Jian, Bisantz, and Drury (2000) are centralized around the
construct of trust, therefore, we see in the network that many of these trust-based
attributes are more related to one another than they are to any car brand. The Nearest
Neighbor Network derived using mean relatedness ratings of each of the brands is shown
in figure 5. In this network we see groups of vehicles emerge, these groups are more
related to one another than the other brands.
Figure 5: Pathfinder Network – Car Brands and Trust Associations (mean ratings, q=n-1, r=inf.)
32
The groups of brands that emerged in figure 6 are solely based on the relatedness
ratings between the twelve brands and twelve trust-based attributes. They were collected
independently of average ranked position. However, a few similarities emerged between
the Nearest Neighbor Network (figure 6) and average ranked position (shown in table 1
and figure 4). Interestingly, the amount of overlap shown in figure 4 between the brand
brackets somewhat matches to the groups that emerged in figure 6. Future work is needed
to determine exactly how related the Nearest Neighbor Network is to average ranked
position (results of the Wilcoxon Signed-Ranked test with a Bonferroni correction) for
each brand. However, one observation that sticks out at first blush is that the top three
brands in table 1 for average ranked position (BMW, Mercedes, Tesla) are the only three
brands to have a median ranked position of 3 of 12. They also most related to each other
in the network, in figure 6.
Correlation
A correlation analysis was used to explore the relationship between brand trust
associations and performance expectations for safety. Brand trust associations were
quantified by number of nodes between each trust-based attribute and each automobile
brand in the pathfinder network. A table for this is shown in (Appendix I). Expected
Figure 6: Nearest Neighbor Network – Car Brands (mean ratings, q=n-1, r=inf,)
33
performance for safety was quantified by the average ranked position from most safe to
least safe (1 to 12), average ranked position for each brand is shown in (table 1). It was
hypothesized that a correlation between brand trust and expected performance would
emerge.
This hypothesis was supported for ten of the twelve trust-based attributes. There
was a significant positive correlation between average ranked position (median) and four
of the trust-based attributes, Confidence (r= 0.693, p< 0.05), Secure (r=0.605, p<0.05),
Integrity (r=0.605, p<0.05), and, Trustworthy (r=0.605, p<0.05). This indicated that, on
average, brands who were more related to these components of trust also tended to be
ranked as more safe.
Additionally, there was a significant negative correlation between average ranked
position (median) and six of the twelve trust-based attributes, Harmful (r=-0.636,
p=0.026), Deceptive (r=-0.636, p=0.026), Underhanded (r=-0.636, p=0.026), Suspicious
(r=-0.636, p=0.026), Beware (r=-0.636, p=0.026) and Familiar (r=-0.732, p=0.007). This
indicates a couple of things. First, on average, brands that were more related to
components of distrust also tended to be ranked as less safe, on average. Additionally, the
brands more related to familiarity also tended to be ranked as less safe.
The correlation between average ranked position (median) and two of the trust-
based attributes was non-significant, Reliable (r=0.15, p=0.642) and Dependable
(r=0.261, p=0.413). These words were more central to the brand-trust association
network and therefore the number of nodes connecting to these attributes did not vary too
much automobile brands.
34
These relationships were significant for both measure of central tendency, mean
and median ranked positions (Appendix J).
35
CHAPTER 5
DISCUSSION
Overview
On average, participants predicted better performance for safety in brands
associated with trust, and worse performance for brands associated with distrust. Results
suggested that brands closely related to the attributes, Confidence, Secure, Integrity, and
Trustworthiness were also expected to produce autonomous vehicle technology that
performs in a safer way. Additionally, brands more related to the attributes Harmful,
Deceptive, Underhanded, Suspicious, Beware and Familiar were also expected to
produce autonomous vehicle technology that performs in a less safe way.
Limitations and Future Work
The present study does bear limitations. For instance, with survey data it is
difficult to ensure that participants were attentive, honest, and appropriately
understanding of the questions. The opportunity for a participant to provide an
explanation for why they make the selections is limited and participants are unable to
clarify or ask questions in they have any. Furthermore, only Mechanical Turk workers are
included in the study therefore it is difficult to determine how well they reflect the
general population of American drivers.
Additionally, in order to limit the length of the survey only of attribute-brand
relatedness ratings were collected for a total of 144 questions in the rating activity.
Therefore, attribute-attribute and brand-brand relatedness ratings had to be derived and
scaled from the original data. It would be interesting to collect all 256 paired comparisons
36
directly in the survey to verify the scaling method in this area of research. Similarly, only
one scenario was used for the ranking activity to collect performance expectations. It
would be interesting to explore performance expectations for safety between brands in
other scenarios besides the pedestrian example. Further work could be done to see if
expectations differ for brands in various situations such as, emergency braking or
adaptive cruise control.
Additionally, individual differences should be further explored. Demographic
information regarding gender, age, education level, personal car ownership, interpersonal
trust, brand preference behavior, and brand trust were collected. Future work should be
done to explore the potential influence of these factors on brand trust with autonomous
vehicle technology and safety. For instance, the sample included 13 Toyota owners, 12
Honda owners, 10 Ford owners, 7 Nissan owners, 5 Chevrolet owners, 5 Mazda owners.
Therefore, 52 of the 99 participant owned cars that made up the bottom middle rank in
expected performance for safety and less associated with trusting attributes.
Furthermore, the Volkswagen brand is an interesting case for further exploration.
This brand had a fairly spread distribution in expected performance for safety and no
participants ranked this brand as most safe (rank 1). Measures of central tendency
describe Volkswagen around the middle level of expected performance for safety.
Though the brand association network illustrates distrusting associations, it would be
interesting to know how many participants were aware that in 2016, Volkswagen was
charged for illegal vehicle software that bypassed standards for diesel emissions
(Boudette, 2017).
37
To further support the findings of the present study, future work should be done
using a similar method. However, instead of collecting relatedness ratings between
brands and trust-based associations, relatedness ratings should be collected for brands and
Aaker’s (1997) empirically derived dimensions of brand personality. This would allow
for a comparison between trust associations and brand personality classifications.
Conclusions
Ultimately, autonomous vehicle system performance will always depend on the
person who is interacting with it, their feelings and willingness to adapt their behavior
and accept the system (Van Geenhuizen & Nijkamp, 2003). Theoretically, perceptions of
trust tend to be based on aspects and expectations for system performance (Lee & See,
2004). Findings from the present study provide limited insight for people’s expectations
level of trustworthiness and performance for different brands of autonomous vehicle
systems. Findings suggest that brand trust associations are related to expected
performance for safety in branded autonomous vehicles.
Trust is important because it implicitly, and sometimes explicitly, informs a
person’s behavior towards a system. People need to make important decisions
surrounding technology that directly impact their own safety, as well as others around
them. If a person over trusts a system, they may rely on it in inappropriate ways. For
instance, they may rely on it for tasks the system is not capable of, or not intended to do,
and this can produce dangerous outcomes. For example, a pedestrian may decide to walk
in front of a vehicle in a situation where the technology is unable to stop for them. The
38
findings of this study show how different brands are associated with different levels and
aspects of trust. These perceptions inform decisions, like whether to walk or not.
Conversely, under trusting the system can also be unsafe. Some systems may
sound an alarm to notify people of various things, like an object in a blind spot or the
need for a person to take over in certain driving conditions. For example, if a person
disregards blind spot alarm and believes they know better than the system, a collision
may occur. Additionally, if a person ignores a signal from the system indicating they
should take control of the automation (drive manually), and they do not take control of
the vehicle, this lack of vigilance and lack of trust in the alarming system can result in an
accident.
This study shows that when compared to other brands, some brands are viewed
with more confidence and trust when it comes to keeping people safe. Therefore, people
may base their behaviors towards a system on their perceptions of the brand. More work
should be done to gain a deeper understanding of these brand differences in trust for
autonomous technology.
Additionally, this work should also encourage producers to be mindful of why
trust in the automation produced by their brand is important. The system’s capabilities
and a person’s expectations for the systems capabilities are important factors to consider,
and they do not always match as well as they should. Safe systems communicate their
capabilities to the person in an appropriate way, and this helps people know how to
interact with the system. Producers should be mindful of this because it may ultimately
affect the safety of their vehicle and its performance. Typically, corporations are focused
39
on portraying their brand in a positive light. However, this study suggests that brand
image and associations are important for more than just sales. This work provides a
unique contribution to the branding literature, because it suggests that brand associations
can impact an autonomous vehicle’s performance and safety on the roadway.
Understanding brand trust associations, how they develop, and how they impact
people’s interactions with technology can help producers create systems in which a
person’s expectations for what the vehicle is capable of matches what the system is
actually capable of. In this way, designing with trust-based expectations in mind may
improve system safety for all parties involved.
40
REFERENCES
Aaker, D. A. (1991). Managing brand equity: capitalizing on the value of a brand name. New York: The Free Press.
Aaker, J. L. (1997). Dimensions of Brand Personality. Journal of Marketing Research, 34(2), 347–356.
Beiker, S. A., & Calo, R. (2012). Legal Aspects of Autonomous Driving. Santa Clara Law Review,52(4), 1145-1156.
Bennett, P. D. (1995). Dictionary of Marketing Terms. Chicago: American Marketing Association.
Boudette, N. E. (2017, November 01). Volkswagen Sales in U.S. Rebound After Diesel Scandal. Retrieved February 10, 2018, from https://www.nytimes.com/2017/11/01/business/volkswagen-sales-diesel.html
Branaghan, R. J., & Hildebrand, E. A. (2011). Brand personality, self-congruity, and preference: A knowledge structures approach. Journal of Consumer Behaviour, 10(5), 304–312.
Carlson, M. S., Desai, M., Drury, J. L., Kwak, H., & Yanco, H. a. (2013). Identifying Factors that Influence Trust in Automated Cars and Medical Diagnosis Systems. The Intersection of Robust Intelligence and Trust in Autonomous Systems: Papers from the AAAI Spring Symposium, (Lin 2008), 20–27.
Choi, J. K., & Ji, Y. G. (2015). Investigating the Importance of Trust on Adopting an Autonomous Vehicle. International Journal of Human-Computer Interaction, 31(10), 692–702.
Deighton, John (1992). The Consumption of Performance. Journal of Consumer Research, 19(3), 362-372.
Delgado-Ballester, E. (2003). Development and validation of a brand trust scale. International Journal of Market Research, 45(1), 35–54.
Desai, M., Medvedev, M., Vázquez, M., Mcsheehy, S., Gadea-Omelchenko, S., Bruggeman, C., Yanco, H. (2012). Effects of changing reliability on trust of robot systems. Proceedings of the seventh annual ACM/IEEE international conference on Human-Robot Interaction - HRI 12.
Fiske, S. T., Cuddy, A. J. C., & Glick, P. (2007). Universal dimensions of social cognition: warmth and competence. Trends in Cognitive Sciences, 11(2), 77–83.
Freling T.H., & Forbes L. P. (2005). An examination of brand personality through methodological triangulation. Journal of Brand Management 13(2), 148–162.
41
Jian, J.-Y., Bisantz, A. M., & Drury, C. G. (2000). Foundations for an Empirically Determined Scale of Trust in Automated Systems. International Journal of Cognitive Ergonomics, 4(1), 53–71.
Keller, K. L. (1993). Conceptualizing, Measuring, and Managing Customer-Based Brand Equity. Journal of Marketing, 57(1), 1-22.
Kotler, P., & Andreasen, A. R. (1991). Strategic marketing for nonprofit organizations. Englewood Cliffs N.J: Prentice-Hall.
Lee, J .D., Moray N. (1992) Trust, control strategies and allocation of function in human- machine systems. Ergonomics, 35, 1243-70.
Lee, J. D., & See, K. a. (2004). Trust in automation: designing for appropriate reliance. Human Factors, 46(1), 50–80.
Mayer, R., Davis, J., & Schoorman, F. (1995). An integration model of organizational trust. Academy of Management Review, 20(3), 709–734.
McCroskey, J. C., & Teven, J. J. (1999). Goodwill: A reexamination of the construct and its measurement. Communication Monographs, 66(1), 90-103
Merritt, S., & Ilgen, D. (2008). Not All Trust Is Created Equal: Dispositional and History-Based Trust in Human Automation Interactions. Human Factors: The Journal of the Human Factors and Ergonomics Society, 50(2), 194–210.
Mosier K, Skitka l, Heers S, B. M. (1998). Automaton Bias: Decision Making and Performance in High-Tech Cockpits. The International Journal of Aviation Psychology, 8(1): 33–45.
Paolacci, G., & Chandler, J. (2014). Inside the Turk: Understanding Mechanical Turk as a participant pool. Current Directions in Psychological Science, 23(3), 184-188.
Parasuraman, R., & Riley, V. (1997). Humans and Automation: Use, Misuse, Disuse, Abuse. Human Factors: The Journal of the Human Factors and Ergonomics Society, 39(2), 230-253.
Peter, J. P., & Ryan, M. J. (1976). An investigation of perceived risk at the brand level. Journal of Marketing Research, 13(2), 184.
Rempel, J.K., Holmes, J.G. & Zanna, M.P. (1985). Trust in close relationships. Journal of Personality and Social Psychology, 49, 95-112.
Robinson, J. P., Shaver, P. R., & Wrightsman, L. S. (1991). Criteria for scale selection and evaluation. Measures of personality and social psychological attitudes, 1(3), 1-16.
42
Rossiter, J. R., & Percy, L. (n.d.). Emotions and Motivations in Advertising. Advances in Consumer Research, 18, 100-110.
Rotter, J. B. (1980). Interpersonal trust, trustworthiness, and gullibility. American Psychologist, 35(1), 1-7.
Schaefer, K. E., Chen, J. Y. C., Szalma, J. L., & Hancock, P. A. (2016). A meta-analysis of factors influencing the development of trust in automation: Implications for understanding autonomy in future systems. Human Factors: The Journal of the Human Factors and Ergonomics Society, 58(3), 377–400.
Schvaneveldt RW. 1990. Pathfinder Associative Networks: Studies in Knowledge Organization. Ablex Publishing: Westport, CT.
Schvaneveldt, R. W., Durso, F. T., & Dearholt, D. W. (1989). Network structures in proximity data. Psychology of Learning and Motivation, 24, 249-284.
Sheridan, T. B. (1988). Trustworthiness of command control systems. Proceedings of the Third IFAC/IFIP/IEA/IFORS Conference on Man Machine Systems (pp. 427-431). Elmsford, NY: Pergamon.
Sheridan, T. B. (1992). Introduction. In Telerobotics, Automation, and Human Supervisory Control (pp. 1-3). Cambridge (Mass.): MIT Press.
Singh I.L, Molloy, R., & Parasuraman, R. (1993). Automation-induced “complacency”: Development of the Complacency-Potential Rating Scale. The International Journal of Aviation Psychology, 3, 111-122.
Thomson, M., MacInnis, D. J., & Whan Park, C. (2005). The Ties That Bind: Measuring the Strength of Consumers’ Emotional Attachments to Brands. Journal of Consumer Psychology, 15(1), 77–91.
United States, National Highway Traffic Safety Administration, U.S. Department of Transportation. (2017). Automated driving systems 2.0: a vision for safety.
Van Geenhuizen, M., & Nijkamp, P. (2003). Coping with Uncertainty in the field of new transport technology. Transportation Planning and Technology. 26(6), 449-467.
Wojciszke, B., & Abele, A. E. (2008). The primacy of communion over agency and its reversals in evaluations. European Journal of Social Psychology, 38(7), 1139-1147.
Zeithaml, V. A. (1988). Consumer Perceptions of Price, Quality, and Value: A Means-End Model and Synthesis of Evidence. Journal of Marketing, 52(3), 2-22.
Example of Likert-style questions for Brand-Trust Association measures (Volvo)
43
APPENDIX A
AMAZON MECHANICAL TURK SURVEY LINK
44
HIT Title: Answer a survey about car brands! HIT Description: This is a survey regarding various car brands and your opinions of them. This study will take about 30 minutes or less to complete. You will be compensated $1.00US for your time and honest participation. You must be at least 18 years old to participate. HIT ID: 324N5FAHSYU0PK45BZJEF2YNGGFKVF
45
APPENDIX B
INFORMED CONSENT
46
47
48
APPENDIX C
RANKING ACTIVITY
49
50
A number appeared next to each brand as it was dragged into the box. The number indicated the brand’s ranked order position (as shown below).
Rank 1 description (shown below)
51
APPENDIX D
RATING ACTIVITY
52
This structure was be repeated for all twelve automobile brands.
53
APPENDIX E
DEMOGRAPHICS QUESTIONAAIRE AND END OF SURVEY
54
1. What is your Gender? • Male • Female • Other • I do not want to provide this information
2. What is your Age? • 18-24 years old • 25-34 years old • 35-44 years old • 45-54 years old • 55-64 years old • 65-74 years old • 75 years or older • I do not want to provide this information
3. Please select the highest level of education you have completed: • No schooling completed • Some high school, no diploma • High school diploma or the equivalent (for example: GED) • Some college, no degree • Trade/technical/vocational training • Associate degree • Bachelor’s degree • Master’s degree • Professional degree • Doctorate degree • I do not wish to provide this information
4. Are you a licensed driver? • Yes • No • I do not wish to provide this information
5. Do you own a car? • Yes • No • I do not wish to provide this information
(ONLY IF “Yes” to previous question)
55
6. What is the make of your primary vehicle (e.g, Jeep, Honda, Audi, Volvo, etc.)?
7. I think most people can be relied on to do what they say they will do. • 1: Strongly Disagree to 7: Strongly Agree
8. I have favorite brands (in general, not just car brands). I prefer their products over
other brands. • 1: Strongly Disagree to 7: Strongly Agree
9. When it comes to my safety, I believe I can rely on some brands more than others.
• 1: Strongly Disagree to 7: Strongly Agree
10. What type of device did you use to complete this survey? • Computer (desktop/laptop) • Mobile Phone • Tablet • Other • I do not wish to provide this information
END OF SURVEY
56
APPENDIX F
FRIEDMAN’S TEST (WITH WILCOXON POST TESTS)
57
58
Wilcoxon Post hoc with Bonferroni correction
59
60
APPENDIX G
RANK DESCRIPTIVE STATISTICS
61
62
Ford
Frequency Percent Valid PercentCumulative
Percent
Valid 1
2
3
4
5
6
7
8
9
10
11
12
Total
Missing System
Total
4 4.0 4.0 4.0
8 8.0 8.1 12.1
2 2.0 2.0 14.1
7 7.0 7.1 21.2
7 7.0 7.1 28.3
4 4.0 4.0 32.3
9 9.0 9.1 41.4
8 8.0 8.1 49.5
14 14.0 14.1 63.6
14 14.0 14.1 77.8
11 11.0 11.1 88.9
11 11.0 11.1 100.0
99 99.0 100.0
1 1.0
100 100.0
Chevrolet
Frequency Percent Valid PercentCumulative
Percent
Valid 1
2
3
4
5
6
7
8
9
10
11
12
Total
Missing System
Total
1 1.0 1.0 1.0
3 3.0 3.0 4.0
5 5.0 5.1 9.1
4 4.0 4.0 13.1
8 8.0 8.1 21.2
8 8.0 8.1 29.3
6 6.0 6.1 35.4
9 9.0 9.1 44.4
14 14.0 14.1 58.6
18 18.0 18.2 76.8
16 16.0 16.2 92.9
7 7.0 7.1 100.0
99 99.0 100.0
1 1.0
100 100.0
Page 6
Chrysler
Frequency Percent Valid PercentCumulative
Percent
Valid 1
2
4
5
6
7
8
9
10
11
12
Total
Missing System
Total
1 1.0 1.0 1.0
1 1.0 1.0 2.0
3 3.0 3.0 5.1
5 5.0 5.1 10.1
7 7.0 7.1 17.2
13 13.0 13.1 30.3
13 13.0 13.1 43.4
9 9.0 9.1 52.5
11 11.0 11.1 63.6
17 17.0 17.2 80.8
19 19.0 19.2 100.0
99 99.0 100.0
1 1.0
100 100.0
Page 7
Nissan
Frequency Percent Valid PercentCumulative
Percent
Valid 1
2
3
4
5
6
7
8
9
10
11
12
Total
Missing System
Total
1 1.0 1.0 1.0
3 3.0 3.0 4.0
4 4.0 4.0 8.1
5 5.0 5.1 13.1
4 4.0 4.0 17.2
13 13.0 13.1 30.3
15 15.0 15.2 45.5
11 11.0 11.1 56.6
15 15.0 15.2 71.7
12 12.0 12.1 83.8
10 10.0 10.1 93.9
6 6.0 6.1 100.0
99 99.0 100.0
1 1.0
100 100.0
Mazda
Frequency Percent Valid PercentCumulative
Percent
Valid 2
3
4
5
6
7
8
9
10
11
12
Total
Missing System
Total
2 2.0 2.0 2.0
2 2.0 2.0 4.0
2 2.0 2.0 6.1
4 4.0 4.0 10.1
11 11.0 11.1 21.2
15 15.0 15.2 36.4
12 12.0 12.1 48.5
11 11.0 11.1 59.6
12 12.0 12.1 71.7
12 12.0 12.1 83.8
16 16.0 16.2 100.0
99 99.0 100.0
1 1.0
100 100.0
Page 5
Frequency Table
TESLA
Frequency Percent Valid PercentCumulative
Percent
Valid 1
2
3
4
5
6
7
8
9
10
11
12
Total
Missing System
Total
37 37.0 37.4 37.4
4 4.0 4.0 41.4
12 12.0 12.1 53.5
9 9.0 9.1 62.6
4 4.0 4.0 66.7
3 3.0 3.0 69.7
4 4.0 4.0 73.7
5 5.0 5.1 78.8
2 2.0 2.0 80.8
1 1.0 1.0 81.8
8 8.0 8.1 89.9
10 10.0 10.1 100.0
99 99.0 100.0
1 1.0
100 100.0
Page 1
63
HONDA
Frequency Percent Valid PercentCumulative
Percent
Valid 1
2
3
4
5
6
7
8
9
10
11
12
Total
Missing System
Total
7 7.0 7.1 7.1
6 6.0 6.1 13.1
11 11.0 11.1 24.2
10 10.0 10.1 34.3
14 14.0 14.1 48.5
12 12.0 12.1 60.6
7 7.0 7.1 67.7
8 8.0 8.1 75.8
4 4.0 4.0 79.8
4 4.0 4.0 83.8
7 7.0 7.1 90.9
9 9.0 9.1 100.0
99 99.0 100.0
1 1.0
100 100.0
Volvo
Frequency Percent Valid PercentCumulative
Percent
Valid 1
2
3
4
5
6
7
8
9
10
11
12
Total
Missing System
Total
15 15.0 15.2 15.2
10 10.0 10.1 25.3
10 10.0 10.1 35.4
15 15.0 15.2 50.5
7 7.0 7.1 57.6
6 6.0 6.1 63.6
6 6.0 6.1 69.7
11 11.0 11.1 80.8
7 7.0 7.1 87.9
6 6.0 6.1 93.9
2 2.0 2.0 96.0
4 4.0 4.0 100.0
99 99.0 100.0
1 1.0
100 100.0
Page 2
Mercedes
Frequency Percent Valid PercentCumulative
Percent
Valid 1
2
3
4
5
6
7
8
9
10
11
12
Total
Missing System
Total
13 13.0 13.1 13.1
20 20.0 20.2 33.3
18 18.0 18.2 51.5
14 14.0 14.1 65.7
11 11.0 11.1 76.8
4 4.0 4.0 80.8
4 4.0 4.0 84.8
2 2.0 2.0 86.9
4 4.0 4.0 90.9
4 4.0 4.0 94.9
3 3.0 3.0 98.0
2 2.0 2.0 100.0
99 99.0 100.0
1 1.0
100 100.0
Volkswagen
Frequency Percent Valid PercentCumulative
Percent
Valid 2
3
4
5
6
7
8
9
10
11
12
Total
Missing System
Total
8 8.0 8.1 8.1
8 8.0 8.1 16.2
11 11.0 11.1 27.3
14 14.0 14.1 41.4
11 11.0 11.1 52.5
13 13.0 13.1 65.7
4 4.0 4.0 69.7
7 7.0 7.1 76.8
6 6.0 6.1 82.8
5 5.0 5.1 87.9
12 12.0 12.1 100.0
99 99.0 100.0
1 1.0
100 100.0
Page 3
64
BMW
Frequency Percent Valid PercentCumulative
Percent
Valid 1
2
3
4
5
6
7
8
9
10
11
12
Total
Missing System
Total
14 14.0 14.1 14.1
24 24.0 24.2 38.4
18 18.0 18.2 56.6
11 11.0 11.1 67.7
8 8.0 8.1 75.8
7 7.0 7.1 82.8
1 1.0 1.0 83.8
6 6.0 6.1 89.9
1 1.0 1.0 90.9
4 4.0 4.0 94.9
3 3.0 3.0 98.0
2 2.0 2.0 100.0
99 99.0 100.0
1 1.0
100 100.0
Toyota
Frequency Percent Valid PercentCumulative
Percent
Valid 1
2
3
4
5
6
7
8
9
10
11
12
Total
Missing System
Total
6 6.0 6.1 6.1
10 10.0 10.1 16.2
9 9.0 9.1 25.3
8 8.0 8.1 33.3
13 13.0 13.1 46.5
13 13.0 13.1 59.6
6 6.0 6.1 65.7
10 10.0 10.1 75.8
11 11.0 11.1 86.9
7 7.0 7.1 93.9
5 5.0 5.1 99.0
1 1.0 1.0 100.0
99 99.0 100.0
1 1.0
100 100.0
Page 4
Toyota
121086420
Frequency
12.5
10.0
7.5
5.0
2.5
0.0
Toyota Mean = 5.92 Std. Dev. = 2.951 N = 99
Nissan
121086420
Frequency
15
10
5
0
Nissan Mean = 7.75 Std. Dev. = 2.639 N = 99
Page 4
Nissan
Frequency Percent Valid PercentCumulative
Percent
Valid 1
2
3
4
5
6
7
8
9
10
11
12
Total
Missing System
Total
1 1.0 1.0 1.0
3 3.0 3.0 4.0
4 4.0 4.0 8.1
5 5.0 5.1 13.1
4 4.0 4.0 17.2
13 13.0 13.1 30.3
15 15.0 15.2 45.5
11 11.0 11.1 56.6
15 15.0 15.2 71.7
12 12.0 12.1 83.8
10 10.0 10.1 93.9
6 6.0 6.1 100.0
99 99.0 100.0
1 1.0
100 100.0
Mazda
Frequency Percent Valid PercentCumulative
Percent
Valid 2
3
4
5
6
7
8
9
10
11
12
Total
Missing System
Total
2 2.0 2.0 2.0
2 2.0 2.0 4.0
2 2.0 2.0 6.1
4 4.0 4.0 10.1
11 11.0 11.1 21.2
15 15.0 15.2 36.4
12 12.0 12.1 48.5
11 11.0 11.1 59.6
12 12.0 12.1 71.7
12 12.0 12.1 83.8
16 16.0 16.2 100.0
99 99.0 100.0
1 1.0
100 100.0
Page 5
Volkswagen
121086420
Frequency
12.5
10.0
7.5
5.0
2.5
0.0
Volkswagen Mean = 6.72 Std. Dev. = 3.117 N = 99
BMW
121086420
Frequency
25
20
15
10
5
0
BMW Mean = 4.07 Std. Dev. = 2.901 N = 99
Page 3
65
Mazda
121086420
Frequency
20
15
10
5
0
Mazda Mean = 8.57 Std. Dev. = 2.556 N = 99
Ford
121086420
Frequency
12.5
10.0
7.5
5.0
2.5
0.0
Ford Mean = 7.67 Std. Dev. = 3.301 N = 99
Page 5
Volkswagen
121086420
Frequency
12.5
10.0
7.5
5.0
2.5
0.0
Volkswagen Mean = 6.72 Std. Dev. = 3.117 N = 99
BMW
121086420
Frequency
25
20
15
10
5
0
BMW Mean = 4.07 Std. Dev. = 2.901 N = 99
Page 3
Chevrolet
121086420
Frequency
20
15
10
5
0
Chevrolet Mean = 8.14 Std. Dev. = 2.85 N = 99
Chrysler
121086420
Frequency
20
15
10
5
0
Chrysler Mean = 8.92 Std. Dev. = 2.566 N = 99
Page 6
66
Histogram
TESLA
121086420
Frequency
40
30
20
10
0
TESLA Mean = 4.64 Std. Dev. = 4.004 N = 99
HONDA
121086420
Frequency
12.5
10.0
7.5
5.0
2.5
0.0
HONDA Mean = 6.14 Std. Dev. = 3.289 N = 99
Page 1
Volvo
121086420
Frequency
15
10
5
0
Volvo Mean = 5.24 Std. Dev. = 3.255 N = 99
Mercedes
121086420
Frequency
20
15
10
5
0
Mercedes Mean = 4.23 Std. Dev. = 2.899 N = 99
Page 2
67
APPENDIX H
RANK 1 DESCRIPTIONS: SUMMARY TABLE
68
Automobile Brand R
ank
1 Fr
eque
ncy
Lea
ders
, 1st
bra
nd in
this
spac
e
te
chno
logy
Adv
ance
d &
inno
vativ
e br
and
in
the
auto
mot
ive
indu
stry
Hig
h Q
ualit
y; L
uxur
y Br
and
Rep
utat
ion
For S
afet
y
Rel
iabl
e Br
and
Fun
ctio
nal B
rand
Tesla 37 12 15 7 3 0 0 Volvo 15 0 0 1 14 0 0 BMW 14 0 2 6 0 6 0 Mercedes 13 0 2 10 0 1 0 Honda 8 0 1 1 0 6 0 Toyota 5 0 0 0 3 2 0 Ford 4 0 0 0 0 4 0 Nissan 1 0 0 0 0 1 0 Chevrolet 1 0 0 0 0 1 0 Chrysler 1 0 0 0 0 0 1 Mazda 0 0 0 0 0 0 0 Volkswagen 0 0 0 0 0 0 0
Total 99 12 20 25 20 21 1
69
APPENDIX I
NUMBER OF NODES BETWEEN EACH BRAND AND EACH ATTRIBUTE
70
Brand
Con
fiden
ce
Trus
twor
thy
Secu
re
Inte
grity
Rel
iabl
e
Dep
enda
ble
Har
mfu
l
Fam
iliar
Dec
eptiv
e
Und
erha
nded
Susp
icio
us
Bew
are
BMW 1 2 3 3 3 4 9 5 7 8 8 9 Mercedes 1 2 3 3 3 4 9 5 7 8 8 9 Tesla 3 2 3 3 1 2 7 3 5 6 6 7 Volvo 4 3 4 4 2 1 8 4 6 7 7 8 Toyota 3 2 3 3 1 2 5 1 3 4 4 5 Honda 5 4 5 5 3 4 5 1 3 4 4 5 Volkswagen 5 4 5 5 3 4 3 1 1 2 2 3 Ford 5 4 5 5 3 4 5 1 3 4 4 5 Nissan 5 4 5 5 3 4 5 1 3 4 4 5 Chevrolet 5 4 5 5 3 4 5 1 3 4 4 5 Mazda 3 2 3 3 1 2 7 3 5 6 6 7 Chrysler 5 4 5 5 3 4 5 1 3 4 4 5
71
APPENDIX J
FULL CORRELATION MATRIX
72
Correlations
Median_Rank Mean_Rank Confidence Trustworth Secure Integrity Reliable Dependable Harmful Familiar Deceptive Underhanded Suspicious Beware
Median_Rank Pearson Correlation
Sig. (2-tailed)
N
Mean_Rank Pearson Correlation
Sig. (2-tailed)
N
Confidence Pearson Correlation
Sig. (2-tailed)
N
Trustworth Pearson Correlation
Sig. (2-tailed)
N
Secure Pearson Correlation
Sig. (2-tailed)
N
Integrity Pearson Correlation
Sig. (2-tailed)
N
Reliable Pearson Correlation
Sig. (2-tailed)
N
Dependable Pearson Correlation
Sig. (2-tailed)
N
Harmful Pearson Correlation
Sig. (2-tailed)
N
Familiar Pearson Correlation
Sig. (2-tailed)
N
Deceptive Pearson Correlation
Sig. (2-tailed)
N
Underhanded Pearson Correlation
Sig. (2-tailed)
N
Suspicious Pearson Correlation
Sig. (2-tailed)
N
Beware Pearson Correlation
Sig. (2-tailed)
N
1 .977** .693* .605* .605* .605* .150 .261 -.636* -.732** -.636* -.636* -.636* -.636*
.000 .013 .037 .037 .037 .642 .413 .026 .007 .026 .026 .026 .026
12 12 12 12 12 12 12 12 12 12 12 12 12 12
.977** 1 .714** .607* .607* .607* .118 .211 -.637* -.708** -.637* -.637* -.637* -.637*
.000 .009 .036 .036 .036 .715 .511 .026 .010 .026 .026 .026 .026
12 12 12 12 12 12 12 12 12 12 12 12 12 12
.693* .714** 1 .901** .901** .901** .278 .194 -.837** -.861** -.837** -.837** -.837** -.837**
.013 .009 .000 .000 .000 .382 .546 .001 .000 .001 .001 .001 .001
12 12 12 12 12 12 12 12 12 12 12 12 12 12
.605* .607* .901** 1 1.000** 1.000** .667* .541 -.732** -.729** -.732** -.732** -.732** -.732**
.037 .036 .000 .000 .000 .018 .069 .007 .007 .007 .007 .007 .007
12 12 12 12 12 12 12 12 12 12 12 12 12 12
.605* .607* .901** 1.000** 1 1.000** .667* .541 -.732** -.729** -.732** -.732** -.732** -.732**
.037 .036 .000 .000 .000 .018 .069 .007 .007 .007 .007 .007 .007
12 12 12 12 12 12 12 12 12 12 12 12 12 12
.605* .607* .901** 1.000** 1.000** 1 .667* .541 -.732** -.729** -.732** -.732** -.732** -.732**
.037 .036 .000 .000 .000 .018 .069 .007 .007 .007 .007 .007 .007
12 12 12 12 12 12 12 12 12 12 12 12 12 12
.150 .118 .278 .667* .667* .667* 1 .865** -.183 -.137 -.183 -.183 -.183 -.183
.642 .715 .382 .018 .018 .018 .000 .568 .671 .568 .568 .568 .568
12 12 12 12 12 12 12 12 12 12 12 12 12 12
.261 .211 .194 .541 .541 .541 .865** 1 -.308 -.277 -.308 -.308 -.308 -.308
.413 .511 .546 .069 .069 .069 .000 .330 .384 .330 .330 .330 .330
12 12 12 12 12 12 12 12 12 12 12 12 12 12
-.636* -.637* -.837** -.732** -.732** -.732** -.183 -.308 1 .955** 1.000** 1.000** 1.000** 1.000**
.026 .026 .001 .007 .007 .007 .568 .330 .000 .000 .000 .000 .000
12 12 12 12 12 12 12 12 12 12 12 12 12 12
-.732** -.708** -.861** -.729** -.729** -.729** -.137 -.277 .955** 1 .955** .955** .955** .955**
.007 .010 .000 .007 .007 .007 .671 .384 .000 .000 .000 .000 .000
12 12 12 12 12 12 12 12 12 12 12 12 12 12
-.636* -.637* -.837** -.732** -.732** -.732** -.183 -.308 1.000** .955** 1 1.000** 1.000** 1.000**
.026 .026 .001 .007 .007 .007 .568 .330 .000 .000 .000 .000 .000
12 12 12 12 12 12 12 12 12 12 12 12 12 12
-.636* -.637* -.837** -.732** -.732** -.732** -.183 -.308 1.000** .955** 1.000** 1 1.000** 1.000**
.026 .026 .001 .007 .007 .007 .568 .330 .000 .000 .000 .000 .000
12 12 12 12 12 12 12 12 12 12 12 12 12 12
-.636* -.637* -.837** -.732** -.732** -.732** -.183 -.308 1.000** .955** 1.000** 1.000** 1 1.000**
.026 .026 .001 .007 .007 .007 .568 .330 .000 .000 .000 .000 .000
12 12 12 12 12 12 12 12 12 12 12 12 12 12
-.636* -.637* -.837** -.732** -.732** -.732** -.183 -.308 1.000** .955** 1.000** 1.000** 1.000** 1
.026 .026 .001 .007 .007 .007 .568 .330 .000 .000 .000 .000 .000
12 12 12 12 12 12 12 12 12 12 12 12 12 12
Correlation is significant at the 0.01 level (2-tailed).**.
*. Page 2Correlation is significant at the 0.05 level (2-tailed).*.
Page 3
top related